| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
| World Community Grid Forums
|
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 2
|
|
| Author |
|
|
Kaoh
Cruncher ROC,Taiwan(NOT PRC) Joined: Nov 20, 2005 Post Count: 18 Status: Offline Project Badges:
|
My Rosetta jobs sometimes stop at certain percentage
----------------------------------------for a long time (ex:40.687% at 6:40pm,7:30pm still at 40.687%.. This is why I called "freeze") When it is restarted,percentage progress soon... Is it normal?Or something wrong? PS:Can I directly reply to the email "wcgrid@us.ibm.com"? I still haven't got any reply.... ![]() [Edit 1 times, last edit by Kaoh at May 31, 2006 4:31:51 PM] |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Hello Kaoh,
----------------------------------------Rosetta updates the percentage when it finishes folding a protein. Each work unit has a preset number of folds required. Normally it is in the range of 120 to 500 folds. Each fold is made on the same protein using a different variable value (starting with a random seed). Often the fold is the same or almost the same from one to the next, but then the value can reach a tipping point and suddenly the protein will fold a very different way. But there is no progress indication while actually folding the protein. If the work unit requires 125 folds, then at the completion of a fold, progress increases by 0.8%. What can happen is that a particular fold might suddenly adopt a configuration that does not converge quickly to a steady state. So it keeps repeating the loop. Eventually the change in each iteration (delta) will fall below the limit (epsilon), indicating that the iterative loop has reached a fairly steady state. This can be a slow process, and there is a loop counter in Rosetta that keeps this from becoming an endless loop. After several hours, depending on computer speed, the partial result is returned to the server together with the information of where a fold becomes 'non-convergent'. It might really be convergent, if we were willing to spend days on that one fold, but life is too short. Usually, the folds proceed at the same, almost unvarying speed, requiring the same number of loops to converge, but occasionally you will see massive changes on a single work unit. There is a heuristic algorithm that estimates how fast this will be when creating a work unit and assigns the number of folds estimated to take about 10 hours (on the average WCG computer). It is surprisingly accurate statistically, as you can see in the Global Statistics for hours per result, but individual work units can differ by an order of magnitude. Lawrence Added: but if it 'freezes' for too long (an amount of time that depends on your computer speed) then assume something glitched on your computer and reboot. If experience indicates that this fixes the problem, then there is something unreliable about your computer - hardware, software, power supply whatever. But if you know about it and how to coax your computer into working - then great! It just means your computer has character! ![]() [Edit 1 times, last edit by Former Member at May 31, 2006 6:40:16 PM] |
||
|
|
|